When Cloud Migrations Go Wrong: Case Studies and the Framework for Getting It Right

When Cloud Migrations Go Wrong: Case Studies and the Framework for Getting It Right Cloud migration is one of the most common largescale technology initiatives in modern organisations. It is also one of the most frequently mismanaged. Gartner has estimated that through the early 2020s, the majority of organisations that attempted cloud migration experienced cost overruns, schedule delays, or failed to achieve the expected benefits. The failures are instructive — not because they reveal exotic technical problems, but because they reveal the same thinking failures that appear in every complex organisational change: insufficient problem definition, unrealistic assumptions, inadequate measurement of the current state, and a failure to account for the human and organisational factors that determine whether technology initiatives succeed. Target Canada: The $7 Billion Lesson in Migration Complexity Target's expansion into Canada in 2013 is one of the most studied retail failures of the modern era. While not a cloud migration per se, it illustrates the failure mode that appears most commonly in largescale technology and operational migrations: the assumption that a system that works in one context can be transplanted to a different context without fundamental redesign. Target's Canadian operation launched with a supply chain and inventory management system that had been adapted from its US operations. The system was not adequately tested at scale before launch. The data that populated it — product dimensions, weights, and inventory counts — was riddled with errors. The result was stores that were simultaneously overstocked in some categories and out of stock in others, with no reliable way to diagnose or fix the problem. Target Canada closed in 2015, after less than two years of operation, at a cost of approximately $7 billion. The postmortem identified the technology migration as a primary contributing factor — specifically, the failure to validate that the system would